Internet of Things (IoT) enabled Data Centres (DC), play a vital role in managing and sustaining modern information-driven infrastructure. Accurate energy prediction is the major requirement of the DC for efficient resource management, management of growing internet infrastructure and increasing demand of digital services. The main challenges in the DC are scalability and the cost effectiveness which are dependent on the accurate energy prediction. Hence there is a need of accurate energy prediction. The work proposes energy prediction model with best agreement between predicted and actual values resulting to approximately zero error and robustness for an IoT enabled DC. The feature normalization concept has been used in energy prediction model to enhance the robustness of the different regression models. Proposed work is validated by comparison with the earlier reported work on energy prediction. Robust Linear Regression-Random Sample Consensus (RLR-RANSAC) and Linear Regression (LR) exhibited remarkable performance with RMSE 6.8176 × 10-17 159(KWh), 7.9565 × 10-17 (KWh) for hourly time span respectively, as compared with earlier reported work. R-squared (R2) values approaching 1 indicated a near-perfect fit to the data. The proposed approach demonstrated overall performance improvement and can be applicable in IoT enabled DC environment.